NExT: Teaching Large Language Models to Reason about Code Execution

arXiv:2404.14662v1 Announce Type: new
Abstract: A fundamental skill among human developers is the ability to understand and reason about program execution. As an example, a programmer can mentally simulate code execution in natural language to debug and repair code (aka. rubber duck debugging). However, large language models (LLMs) of code are typically trained on the surface textual form of programs, thus may lack a semantic understanding of how programs execute at run-time. To address this issue, we propose NExT, a …

abstract arxiv code cs.cl cs.lg cs.pl cs.se debug debugging developers example fundamental however human language language models large language large language models llms natural natural language next programmer reason repair teaching type